Intracellular Binding Site for a Positive Allosteric Modulator of the Dopamine D1 Receptor

The binding site for DETQ [2-(2,6-dichlorophenyl)-1-((1S,3R)-3-(hydroxymethyl)-5-(2-hydroxypropan-2-yl)-1-methyl-3,4-dihydroisoquinolin-2(1H)-yl)ethan-1-one], a positive allosteric modulator (PAM) of the dopamine D1 receptor, was identified and compared with the binding site for CID 2886111 [N-(6-tert-butyl-3-carbamoyl-4,5,6,7-tetrahydro-1-benzothiophen-2-yl)pyridine-4-carboxamide], a reference D1 PAM. From D1/D5 chimeras, the site responsible for potentiation by DETQ of the increase in cAMP in response to dopamine was narrowed down to the N-terminal intracellular quadrant of the receptor; arginine-130 in intracellular loop 2 (IC2) was then identified as a critical amino acid based on a human/rat species difference. Confirming the importance of IC2, a β2-adrenergic receptor construct in which the IC2 region was replaced with its D1 counterpart gained the ability to respond to DETQ. A homology model was built from the agonist-state β2-receptor structure, and DETQ was found to dock to a cleft created by IC2 and adjacent portions of transmembrane helices 3 and 4 (TM3 and TM4). When residues modeled as pointing into the cleft were mutated to alanine, large reductions in the potency of DETQ were found for Val119 and Trp123 (flanking the conserved DRY sequence in TM3), Arg130 (located in IC2), and Leu143 (TM4). The D1/D5 difference was found to reside in Ala139; changing this residue to methionine as in the D5 receptor reduced the potency of DETQ by approximately 1000-fold. None of these mutations affected the activity of CID 2886111, indicating that it binds to a different allosteric site. When combined, DETQ and CID 2886111 elicited a supra-additive response in the absence of dopamine, implying that both PAMs can bind to the D1 receptor simultaneously.

PAMs of GPCRs have been known for over 2 decades (Bruns and Fergus, 1990;Nemeth et al., 1998), but only recently has the diversity of potential allosteric sites on these receptors been recognized (Congreve et al., 2017). For class A GPCRs, the most well documented site for PAMs and NAMs is the vestibule (Kruse et al., 2013), the site on the extracellular face between extracellular loops 2 and 3 through which orthosteric ligands must pass before entering the deeper orthosteric site situated between the transmembrane (TM) helices. In class C GPCRs, whose orthosteric sites are located in a separate extracellular domain, allosteric sites are often located in the interior of the TM barrel in roughly the same location as the orthosteric site in class A GPCRs (Conn et al., 2014).
Finally, the dopamine D1 PAM "compound B" was shown by site-directed mutagenesis to bind to a cleft in intracellular loop 2 (IC2) (Lewis et al., 2015), a part of the receptor involved in receptor activation and G-protein coupling. In this study, we find that the D1 PAM DETQ also binds to this site. Using chimeric receptors and mutation of individual amino acids, we identify residues important for activity of DETQ at the D1 receptor and for selectivity versus the closely related D5 and b2 receptors. As a comparator, we also studied CID 2886111 [N-(6-tert-butyl-3-carbamoyl-4,5,6,7-tetrahydro-1-benzothiophen-2-yl)pyridine-4-carboxamide], a D1 PAM from a series discovered by the Sibley group (Luderman et al., 2016). We find that CID 2886111 is unaffected by alterations to IC2, indicating that it binds to a different, as yet unidentified, site. Interestingly, although DETQ and CID 2886111 separately only have slight allo-agonist activity, the combination of the two in the absence of dopamine produces a much larger cAMP response than either PAM alone, as is predicted if both PAMs stabilize the same activated conformation by binding to separate sites. These results imply the presence of multiple allosteric sites on the D1 receptor and therefore multiple opportunities for discovery of allosteric modulators of GPCRs.

Materials and Methods
Materials. DETQ was synthesized as previously described (Beadle et al., 2014 The various mutants and chimeras were generated either by polymerase chain reaction-based mutagenesis using the above wild-type cDNA clones as templates or by gene synthesis at GenScript (Piscataway, NJ). The nucleotide sequences encoding full-length wild-type, mutants, and chimeras were inserted into pcDNA3.1hyg or pJTI R4 CMV-TO (Life Technologies, Carlsbad, CA) and verified by DNA sequencing.
Switchover points for all chimeras are described in Supplemental  Table 1.
Protocols for Receptor Expression. For transient transfection, wild-type receptors and mutants were generated by polymerase chain reaction and chimeras were created by DNA synthesis. DNA was then cloned into the pcDNA3.1 vector and was transiently transfected using FuGENE HD (Promega, Madison, WI) into human embryonic kidney 293 (HEK293) cells. Transfected cells were cultured in Dulbecco's modified Eagle's medium with high glucose supplemented with 5% heat inactivated, dialyzed fetal bovine serum (FBS), 1 mM sodium pyruvate, 20 mM HEPES, and 2 mM L-glutamine at 37°C in an atmosphere containing 5% CO 2 for 48 hours. Cells were harvested and suspended in freeze media [FBS with 6% dimethylsulfoxide (DMSO)] at 10 7 cells/ml, and aliquots were stored in liquid nitrogen.
Measurement of cAMP Response. For experiments in stable cell lines, DETQ and CID 2886111 were diluted in DMSO and dispensed into assay plates (ProxiPlate-384 Plus; PerkinElmer, Waltham, MA) using acoustic dispensing (ECHO; Labcyte, San Jose, CA). To each well containing compound or DMSO blank was added 5 ml STIM buffer (Hanks' balanced salt solution supplemented with 0.1% bovine serum albumin, 20 mM HEPES, 500 mM 3-isobutyl-1-methylxanthine, and 100 mM ascorbic acid) containing a 2Â EC 20 concentration of dopamine, followed by cells (2000 cells/well) in 5 ml STIM. The final DMSO concentration was 0.8%. Plates were incubated at room temperature for a total reaction time of 60 minutes. cAMP production was quantified using homogeneous time-resolved fluorescence detection (Cisbio, Bedford, MA) according to the vendor instructions: lysis buffer containing anti-cAMP cryptate (5 ml) and D2-conjugate (5 ml) was added to the wells, plates were incubated for an additional 60-90 minutes, and timeresolved fluorescence was detected using an EnVision plate reader (PerkinElmer). Experiments in transiently transfected cells were carried out as described above except that each well contained 6000 cells, all aqueous additions were in a volume of 10 ml, the final volume of the incubation was 20 ml, dilutions were carried out with an automated pipetting station, and the final compound dispensing step used a Pin Tool (Hamilton, Reno, NV) (100 nl volume). Intracellular Binding Site for a Dopamine D1 PAM 1233 at ASPET Journals on January 9, 2021 molpharm.aspetjournals.org Fluorescence data were converted to cAMP concentrations using a cAMP standard curve. For potentiator-mode concentration-response curves (CRCs), results for each construct were expressed as the percentage of the window between an EC 20 concentration of dopamine alone and the maximum response to dopamine in that construct. This normalization was carried out separately for each plate, and individual data points from six or more plates (representing replicates from at least 3 separate days) were merged into a single GraphPad data table (GraphPad Inc., La Jolla, CA) for each experiment. The potency of dopamine varied up to 60-fold between different constructs, presumably due to effects of the mutations on coupling or expression. For this reason, the EC 20 concentration of dopamine was determined separately for each construct (Supplemental Fig. 1; Supplemental Tables 2-4).
The Jump-In system integrates the gene to be expressed at a single site that is controlled by a tetracycline-inducible cytomegalovirus promotor, resulting in high expression. B max values for wild-type D1, the V119A mutant, and the W123A mutant in 3H-SCH23390 (7-chloro-3methyl-1-phenyl-1,2,4,5-tetrahydro-3-benzazepin-8-ol) binding were 6.0 6 0.3, 8.3 6 0.2, and 4.7 6 0.2 pmol/mg protein, respectively, compared with 0.36 6 0.02 for the hD1 cell line used in the original characterization of DETQ (Svensson et al., 2017). Although allo-agonist activity of DETQ is greater in the high-expression Jump-In D1 line, potentiator activity of DETQ is essentially the same regardless of receptor expression level (Wang and Heinz, unpublished data), in agreement with the conclusions from a previous study of a series of metabotropic glutamate receptor 5 PAMs (Noetzel et al., 2012).
Curve-Fitting Analysis. For each construct/PAM combination, a single curve was fit to data that were normalized and merged as described above. cAMP values were initially fit to a four-parameter logistic equation using GraphPad software (version 7). Fitted bottom values were consistently found to fall within the range of 62% in all mutant and chimera experiments; based on this result, the bottom was fixed to 0% for final curve generation and analysis. In the experiment investigating interactions between the two PAMs (see Fig. 5), the bottom of the CRC for one PAM depended on the concentration of the other, and the bottom was therefore allowed to vary freely in this analysis.
The S.E. for the best-fit value of each curve-fitting parameter was calculated as described in the GraphPad 7 Curve Fitting Guide: where P i is the ith parameter, SS is the sum of squared residuals, DF is degrees of freedom (number of data points minus number of fitted parameters), and Cov(i,i) is the ith diagonal term of the covariance matrix.
The S.E. of the log EC 50 provided by GraphPad was converted to the S.E. of the untransformed (linear) EC 50 by the following equation: S:E: linear 5 lnð10Þ • EC 50 • S:E: log: Fitting of Incomplete Curves. In a few constructs in which potency of the PAM was very low, only the initial rising part of the curve was measurable. If the data points do not begin to turn down into a plateau, the relative activity ratio (RA) (see the Results) is still well defined, since it depends only on the initial slope; however, the EC 50 and top cannot be separately determined, since any two values in the same ratio will fit the same initial slope. In this situation, a twostep procedure was followed to achieve a stable fit. First, the Hill coefficient was fixed to 1; this was supported by the observation that fully defined curves consistently had Hill slopes around 1 (Supplemental Table 2). For three data sets that showed a small degree of downturn at higher concentrations, this resulted in well defined EC 50 and top values (see Figs. 7 and 8). For two other data sets (see Fig. 8), the Hill coefficient was fixed to 1 and the top was fixed to 100%, providing a well defined EC 50 value for calculation of RA.
Construction of a Homology Model Based on an Agonist-State Crystal Structure of the b2-Adrenergic Receptor. The b2 agonist-state crystal structure 3p0g (Rasmussen et al., 2011a) was imported into the Prime module of the Schrödinger software suite (2011 version; Schrödinger Software, New York, NY) and the ligand and nano-antibody structures were deleted. The human D1 receptor sequence was aligned with the b2 sequence and a homology model was constructed using the Structure Prediction Wizard in Prime. There were no insertions or deletions in TM3, IC2, or TM4. The possibility that the IC2 loop could function as a ligand binding site was confirmed using Schrödinger SiteMap.
A simplified analog of DETQ with the 3-and 5-position groups deleted was docked into the IC2 cleft using Schrödinger Glide. In initial docking poses, the ligand consistently adopted a conformation in which the dichlorophenyl ring was nearly coplanar with the tetrahydroisoquinoline ring. In contrast, studies of the ligand alone indicated a strong energetic preference for the dichlorophenyl ring to be nearly perpendicular to the tetrahydroisoquinoline ring, with a prohibitive energetic penalty for coplanarity. Based on this result, the ligand was redocked in its low-energy conformation using the Schrödinger induced-fit protocol with flexible protein and rigid ligand. The ligand was able to fit into the IC2 cleft in several different orientations, and the final pose (see Supplemental Fig. 2 for the Protein Data Bank file) was chosen for its ability to accommodate substitution at the 3-and 5-positions, in agreement with known structure-activity relationships (Beadle et al., 2014). Finally, the 3-and 5-position groups of DETQ were added to the docked structure and the protein-ligand complex was optimized using Prime.

D1 PAMs.
The main purpose of this study was to identify and characterize the binding site for DETQ (Fig. 1), a novel D1 PAM from a series of acyl-tetrahydroisoquinolines (Beadle et al., 2014;Svensson et al., 2017;Bruns et al., 2018). DETQ is a potent allosteric potentiator at the human D1 receptor (EC 50 , 5.8 nM) with 30-fold lower activity at the rat D1 receptor and more than 1000-fold lower activity at the human D5 receptor.
In a study of this kind, it would be helpful to have a comparator compound from an unrelated chemical series, preferably one that bound to a different site. We have used CID 2886111 ( Fig. 1) for this purpose. We identified CID 2886111 from its close structural similarity to CID 2862078 [6-tert-butyl-2-(thiophene-2-carbonylamino)-4,5,6,7-tetrahydro-1-benzothiophene-3-carboxamide], which was reported to be active in a D1 PAM assay by the Sibley group at the National Institutes of Health (https://pubchem.ncbi.nlm.nih.gov/bioassay/ 504651#section5Data- Table) (Luderman et al., 2016). As described below, CID 2886111 binds to a separate site from DETQ, and functional data imply that DETQ and CID 2886111 can bind to the D1 receptor simultaneously.
D1/D5 Chimeras. A first step toward exploring the binding site for DETQ would be to identify its approximate location on the D1 receptor. Based on the .1000-fold preference of at ASPET Journals on January 9, 2021 molpharm.aspetjournals.org DETQ for the D1 receptor over the D5, we replaced regions of the D1 receptor with their D5 counterparts (unless otherwise stated, all D1, D5, and b2 receptor constructs refer to the human sequences). By switching out large domains, it should be possible to narrow down the binding site without any prior knowledge of its location. Four chimeras were designed, each replacing about half of the D1 receptor with its D5 counterpart. The first two replaced either the N-terminal or C-terminal half of the D1 receptor with the D5 sequence, with the dividing line located between His164 4.66 and Lys165 4.67 at the C-terminal end of TM4 [see Ballesteros and Weinstein (1995) for the residue numbering convention; following the GPCRdb database, we define the last residue of TM3 as Ser127 3.56 and the first residue of TM4 as Thr136 4.38 ]. Two other chimeras replaced either the extracellular or intracellular half of the D1 receptor with the D5 sequence, with the seven switchover points occurring in the middle of each TM segment (see Supplemental Table 1 for the exact locations of the switchover points). Finally, to identify vestibule binders, two additional chimeras swapped out only extracellular loop 2, leaving the rest of the receptor either D1 or D5.
For each construct, a CRC of each PAM for accumulation of cAMP was carried out in the presence of an EC 20 concentration of dopamine (Fig. 2). The RA (Ehlert, 2005;Kenakin, 2017), calculated as the fitted top divided by the EC 50 value, was used as a single measure of potency. If the Hill coefficient is near 1, as seen for the majority of curves in this study (Supplemental Table 2), RA is equivalent to the initial slope of the CRC when plotted on a linear scale. The effect of an experimental intervention such as receptor mutagenesis is conveniently expressed as intrinsic RA (Ehlert, 2005), defined in our study as RA of the mutant construct as a percentage of the RA for the wild-type receptor.
In this study, DETQ was about 1000-fold less potent at the D5 receptor than at the D1. The constructs in which the N-terminal half or the intracellular half of the D1 receptor were replaced with their D5 counterparts showed a similar loss of affinity for DETQ, whereas the other two half-chimeras showed activity similar to wild-type D1. These results indicate that the binding site for DETQ is in the N-terminal intracellular portion of the receptor.
From the above information, it is possible to deduce the amino acid responsible for the human/rat affinity difference, and hence the location of the binding site for DETQ. The only amino acid in the N-terminal intracellular portion of the receptor that differs between rat and human is arginine-130 (Arg130 IC2.3 ) (Monsma et al., 1990;Zhou et al., 1990), implying that the binding site is located in IC2. This location was previously reported as the binding site for the D1 PAM "compound B" (Lewis et al., 2015). This finding is also in agreement with results of human/rat chimera studies carried out at Lilly early in the D1 project (Gadski, Beavers, Little, Yang, and Bruns, unpublished data). Experiments confirming that an R130Q mutation accounts for the human/rat species difference are described below.
CID 2886111 had nearly the same affinity at the D5 receptor as the D1, although the maximum D5 response was only about  Table 2. one-third of the D1 response (Fig. 2). The resultant shift in RA of only 3-fold was insufficient to distinguish robustly between D1-like and D5-like activity, and the results with the chimeras were ambiguous.
Evidence That Mutation of Arginine-130 to Glutamine in IC2 Is Responsible for the Human/Rat Species Difference in Potency of DETQ. To confirm that arginine-130 was responsible for the human/rat species difference, we mutated this residue to glutamine (the amino acid present in rat) and also created the reverse mutation (Q129R in the rat sequence). The R130Q mutation shifted the human receptor to a rat-like potency and the Q129R mutation of the rat D1 receptor reversed this shift, confirming that Arg130 IC2.3 is responsible for the human/rat species difference (Fig. 3). Inspection of published D1 sequences in the UniProt database (http://www.uniprot.org) shows that arginine is ancestral and the mutation to glutamine occurred in the crown of the rodent line, since the rat, mouse, and guinea pig show the glutamine mutation, whereas the rabbit (in Lagomorpha, a sister order to Rodentia) retains arginine, as do distant species such as Xenopus and Drosophila.
CID 2886111 had 2.4-fold higher potency at the rat D1 receptor compared with human D1 and was unaffected by the R130Q mutation. The divergent behavior of CID 2886111 compared with DETQ hints that their binding sites may be different (see below).
Insertion of the D1 IC2 Region into the b2-Receptor Confers Sensitivity to DETQ. The b2-adrenergic receptor, although closely related to the D1 receptor, does not respond to DETQ (Svensson et al., 2017). To find out whether the IC2 region is responsible for the PAM activity of DETQ, we replaced this region of the b2 receptor with the Fig. 4. Substitution of the D1 IC2 region into the b2-adrenergic receptor confers sensitivity to DETQ. The D1 wild-type construct was tested in the presence of an EC 20 concentration of dopamine and the two b2 constructs were tested with an EC 20 concentration of norepinephrine. All experiments were carried out using stable cell lines. Values are best-fit parameters 6 S.E. (n = 8) from nonlinear least-squares curve-fitting to a four-parameter model with the bottom of the dopamine or norepinephrine window fixed to zero. Additional details including Hill coefficients and EC 20 dopamine or norepinephrine concentrations are provided in Supplemental Table 2.
Intracellular Binding Site for a Dopamine D1 PAM 1237 at ASPET Journals on January 9, 2021 molpharm.aspetjournals.org corresponding region from the D1 receptor. The residues that were replaced, consisting of IC2 and adjacent parts of TM3 and TM4 (V 3.45 through I 4.46 ), were chosen based on a homology model described below. DETQ robustly potentiated the response to norepinephrine in this construct, with a potency about 5-fold lower than at the human D1 receptor (Fig. 4). This result indicates that the IC2 region is sufficient to confer PAM activity of DETQ, although the 5-fold loss of potency suggests an auxiliary role for residues outside this region.
CID 2886111 blocked the activation of the b2 receptor by norepinephrine (Fig. 4). This result suggests that CID 2886111 may be a NAM at the b2 receptor, a possibility that should be investigated in more detail. Replacing the b2 IC2 region with the corresponding D1 sequence did not restore PAM activity, indicating that CID 2886111 binds to a different site than DETQ.
Evidence from Interaction Studies That DETQ and CID 2886111 Bind to Different Sites. If DETQ and CID 2886111 bind to different sites but stabilize the same receptor conformation, the theory of linked equilibria (Monod et al., 1965;Koshland et al., 1966;Canals et al., 2012;Changeux and Christopoulos, 2017) predicts that they will act cooperatively (supra-additively) to activate the receptor. In the absence of dopamine, each PAM by itself increased cAMP to only about 2% of the dopamine maximum, but in combination (without  (Ehlert, 2005;Griffin et al., 2007;Kenakin, 2014) with the a (affinity) and b (efficacy) parameters each fitted and graphed separately. Allosteric parameters were as follows: 2.2 6 0.1 for a and 13.2 6 1.6 for b for CID 2886111 modulating the CRC for DETQ (C), and 1.91 6 0.12 for a and 9.0 6 2.2 for b for DETQ modulating the CRC for CID 2886111 (D). The stable human D1 cell line used in this experiment is described in Svensson et al. (2017).  at ASPET Journals on January 9, 2021 molpharm.aspetjournals.org Downloaded from dopamine) they increased cAMP to about 23% of the dopamine maximum (Fig. 5), a much higher response than predicted by additivity. Each PAM also shifted the EC 50 of the other about 2-fold to the left. In contrast to the synergy between DETQ and CID 2886111, combinations of two PAMs from the acyltetrahydroisoquinoline series (Beadle et al., 2014) did not produce a response higher than the maximum for either compound tested separately (unpublished observation). The mutual synergy of dopamine, DETQ, and CID 2886111 implies that they bind to three separate sites yet drive the same receptor conformation. A recent abstract provides similar evidence for two separate D1 PAM sites (Luderman et al., 2018).
Homology Model of the Intracellular Binding Site for DETQ. To explore the binding site for DETQ, a homology model was constructed based on 3p0g, an agonist-state crystal structure of the human b2 receptor in complex with a nano-antibody Downloaded from (Rasmussen et al., 2011a). Confidence in the model was supported by the high homology between the D1 and b2 receptors (14 of 28 amino acids identical in the IC2 region, defined as V116 3.45 through L143 4.45 ) and the lack of any insertions or deletions in the IC2 region. In the b2 receptor and the homology model, this region consists of a twisted loop, with TM3 passing under TM4 and IC2 connecting the two in a retrograde direction compared with the other two IC loops (Fig. 6). The middle of the IC2 bend is organized into a short a-helix. Notably, the inside surface of the bend forms a cleft large enough to accommodate a small-molecule ligand such as DETQ. Unlike the outer surface of the IC2 bend (Rasmussen et al., 2011b), the inner surface has few direct interactions with G s or neighboring TM segments and is relatively less conserved (Ballesteros and Weinstein, 1995), potentially explaining the high specificity of DETQ for the D1 receptor over closely related receptors. Residues that line the inside of the loop (and therefore may interact with a small-molecule PAM ligand) are labeled in Fig. 6.
DETQ was able to dock in several orientations, and the pose shown was selected based on consistency with observed structure-activity relationships (Beadle et al., 2014). The nearly flat tetrahydroisoquinoline ring system lies across the cleft, with the benzene ring of the tetrahydroisoquinoline at the "upper" (intramembrane) end of the cleft. The dichlorophenyl moiety is oriented nearly perpendicular to the tetrahydroisoquinoline ring and is sandwiched between the side chains of W123 3.52 and R130 IC2.3 . The amide group forms a bridge between the two ring systems, forming a potential hydrogen bond with the amino group of K134 IC2.4 . The 3-position hydroxyl forms a potential hydrogen bond with the carbonyl of K134.
The "right-hand" or TM3 side of the cleft is lined by V116 3.45 , V119 3.48 , and W123 3.52 , the latter two residues bracketing the canonical DRY sequence that is known to be involved in agonist coupling (Rasmussen et al., 2011a,b). R130 IC2.3 and K134 IC2.4 line the inner surface of the a-helical portion of IC2, whereas A139 4.41 and L143 4.45 are located on the left side of the cleft, although A139 is recessed and does not interact with DETQ in the model. Finally, side chains of Y131 IC2.4 and M135 IC2.8 are located on the floor of the cleft along with the benzene ring of F62 2,42 , which intercalates into the IC2 loop from TM2.
Alanine Scan of the Binding Site. The amino acids modeled as forming potential contacts with DETQ were each mutated to alanine. CID 2886111, which binds to a different site, was used as a control to monitor possible effects of changes in receptor expression or coupling in the mutants that might masquerade as disruption of binding contacts with DETQ.
The W123 3.52 A mutation caused the largest effect, roughly a 500-fold loss of potency (Fig. 7). In agreement with its importance, the large planar ring system is modeled as extending parallel to the cleft, forming a large portion of the surface area of the TM3 side of the cleft. The electron-deficient dichlorophenyl group forms a p-stacking interaction with the electron-rich tryptophan ring in the homology model. Interestingly, the V119 3.48 A replacement lowered efficacy of DETQ to about 15% without affecting the EC 50 . Mutation of R130, K134, M135, and L143 to alanine also had substantial effects on potency, ranging from about 4-fold for K134 to about 60-fold for R130. The large effects of these residues confirm the general topology of the IC2 region in the homology model. The relatively weak (4-fold) effect of the K134A mutation suggests that the H-bond between the K134 amine and the bridge carbonyl of DETQ in the homology model may not exist or may be energetically unimportant.
In the activated state of the b2 receptor, Y131 IC2.4 forms a hydrogen bond with the aspartate in the DRY sequence (Rasmussen et al., 2011b). To separate the role of this residue in coupling from its potential role in forming a binding interaction with DETQ, we created two mutants. In Y131F, tyrosine is replaced with phenylalanine, which lacks hydrogenbonding ability but retains the same aromatic ring, whereas in Y131A the alanine lacks both hydrogen-bonding and  at ASPET Journals on January 9, 2021 molpharm.aspetjournals.org Downloaded from aromaticity. As expected, the functional affinity of dopamine was considerably weaker (about 50-fold) in both mutants (Supplemental Fig. 1; Supplemental Tables 3 and 4). Additionally, in both mutants, the combination of DETQ and dopamine was able to elicit a much larger maximum response than dopamine alone, indicating that dopamine is incapable of fully activating the receptor by itself in these mutants.
In agreement with this interpretation, the maximum response to dopamine in these two mutants was considerably less than that seen in wild-type D1 (Supplemental Fig. 1; Supplemental Tables 3 and 4). When RA was calculated relative to the dopamine maximum in the same mutant, the Y131A mutant lowered RA by about 13-fold, whereas the Y131F mutant reduced RA by only 3-fold. This result suggests that the alanine at ASPET Journals on January 9, 2021 molpharm.aspetjournals.org substitution disrupted a specific interaction between DETQ and its binding site, whereas the phenylalanine substitution disrupted general coupling of the D1 receptor but had a smaller effect on its specific binding interaction with DETQ. A similar pattern was seen with the F62 2.42 A mutation, in which the potency of dopamine was decreased about 15-fold (Supplemental Table 3), but the RA of DETQ was reduced only 3-fold (Fig. 7), indicating a strong effect on coupling or expression but only a modest effect on affinity of DETQ.
Of the residues modeled as potentially interacting with DETQ, V116 3.45 was the only one to lack any measurable effect when replaced with alanine (Fig. 7). This result constrains how far the DETQ molecule extends up the IC2 cleft.
The V58 2.38 A mutation was previously reported to lower the affinity of compound B by 220-fold (Lewis et al., 2015). In our homology model, the side chain of this residue does not extend into the PAM binding site, but does underlie Met135 IC2.8 , leaving open the possibility of an indirect influence on PAM binding. However, theV58A mutation had no effect on potency of DETQ, suggesting a difference in binding mode between DETQ and the more bulky compound B.
The potency of CID 2886111 was unaffected by any of the above mutations, confirming that the effects on potency of DETQ seen with the mutants reflect disruption of specific binding interactions with DETQ as opposed to secondary effects on other parameters such as receptor expression or coupling.
Role of Alanine-139 in the D1/D5 Species Difference in Affinity for DETQ. The alanine scan described above did not identify the residue responsible for the roughly 1000-fold loss of affinity of DETQ at the D5 receptor. We therefore separately mutated each residue in the D1 IC2 region that differed from its D5 counterpart (Fig. 8).
Seven of the eight residues tested either had little effect on potency of DETQ or (in the case of L143 4.45 M) actually increased potency. However, the A139 4.41 M mutation caused a striking .1000-fold loss of potency of DETQ. Interestingly, in the homology model, the side chain of alanine-139 is recessed and does not appear to interact with DETQ, but its small size leaves open a niche in the side of the binding cleft into which the 3-position hydroxymethyl of DETQ extends. The much larger methionine side chain present in the D5 receptor would cause severe steric interference with the 3-hydroxymethyl of DETQ in this binding pose. As expected, replacement of methionine with alanine in this position of the D5 receptor restored affinity for DETQ (Fig. 8).
None of the IC2 mutations had any notable effect on activity of CID 2886111.
Amino Acids Involved in the D1/b2 Difference in Affinity for DETQ. The origin of DETQ's inactivity at the b2-adrenergic receptor was also of interest. The four residues that interacted with DETQ in the homology model and differed between the D1 and b2 receptors were each separately mutated to their b2 counterparts. (Alanine-139, the amino acid involved in the D1/D5 difference, is altered to lysine in the b2 receptor; the effect of this change was not investigated.) Each of the substitutions caused a substantial loss of potency, ranging from 7-to 80-fold (Fig. 9). The 80-fold loss of potency with the W123F mutation was less severe than the 500-fold loss with the W123A mutation (Fig. 7), and the same was true for R130K compared with R130A (7-fold and 60-fold, respectively). The other two mutants (K134L and M135L) both showed roughly the same magnitude of effect as the corresponding alanine substitutions. None of the mutations affected the potency of CID 2886111.

Discussion
This report describes two sites for allosteric modulation of the dopamine D1 receptor. Results with D1 receptor chimeras and mutants show that the D1 PAM DETQ occupies a cleft in IC2 as previously described for compound B (Lewis et al., 2015). In addition, we show that another D1 PAM, CID 2886111, occupies a separate site (exact location not yet identified) that interacts synergistically with the DETQ site. These results imply a rich structural landscape for allosteric modulation of the D1 receptor and other GPCRs. at ASPET Journals on January 9, 2021 molpharm.aspetjournals.org

Downloaded from
The IC2 region to which DETQ binds is critically involved in the structural changes that accompany the transition between the inactive and active conformations of class A GPCRs. Upon receptor activation, Asp 3.49 of the conserved DRY sequence at the cytoplasmic end of TM3 breaks a hydrogen bond with Arg 3.50 and forms a new hydrogen bond with Tyr IC2.4 , freeing Arg 3.50 to interact with Tyr391 of the a-subunit of G s (Rasmussen et al., 2011b). In this study, Val119 3.48 and Trp123 3.52 , which flank the DRY sequence in the D1 receptor, were found to be critical for DETQ potency. Interestingly, the V119A mutation reduced the efficacy of DETQ to about 15%, suggesting that DETQ may exert a steric push on Val119, which in turn may rotate the DRY sequence and facilitate a remodeling of its hydrogen-bond network into the activated-state pattern. In a similar light, Phe IC2.2 , immediately adjacent to R130 IC2.3 , binds to a hydrophobic pocket in the a-subunit of G s and has a critical role in G-protein coupling (Moro et al., 1993;Rasmussen et al., 2011b). These results taken together provide hints with respect to how DETQ enhances coupling of the D1 receptor to G s .
The intracellular location of the IC2 site may have implications for the pharmacology of DETQ and its congeners. The acyl-tetrahydroisoquinolines (Beadle et al., 2014) are quite hydrophobic and generally should cross cell membranes easily, but the intracellular location of the site should be borne in mind when interpreting structure-activity relationships based on whole-cell assays. The lack of probe dependence seen with DETQ (Svensson et al., 2017) may relate to the relatively large distance of the IC2 site from the orthosteric Fig. 9. Replacement of D1 residues with their b2 counterparts. CRCs of DETQ and CID 2886111 for accumulation of cAMP were carried out in the presence of an EC 20 concentration of dopamine. All experiments were carried out using transient expression. Values are best-fit parameters 6 S.E. (n = 6) from nonlinear least-squares curve-fitting to a four-parameter model with the bottom of the dopamine window fixed to zero. Additional details including Hill coefficients and EC 20 dopamine concentrations are provided in Supplemental Table 2.
Intracellular Binding Site for a Dopamine D1 PAM site; in contrast, muscarinic PAMs and NAMs that bind to the vestibule can show strikingly different effects in modulating the affinity of different orthosteric ligands, both agonists and antagonists (Stockton et al., 1983;Valant et al., 2012). This may be related to the vestibule's position immediately adjacent to the orthosteric site, so that binding to the vestibule can cause local changes in the shape of the orthosteric site and vice versa (Kruse et al., 2013). In addition to the above, some vestibule binders block the entry and exit of orthosteric ligands, slowing kinetics and mandating that the orthosteric ligand bind before the allosteric ligand (Proska and Tucek, 1994). These characteristics would not necessarily pertain to allosteric modulators that bind to the IC2 site or other sites that are more remote from the orthosteric site. It is possible that IC2 binders could exert local effects on the adjacent G protein, which would likely manifest as selectivity for the G-protein signaling pathway. DETQ has roughly equal potency for the G s and b-arrestin pathways (Svensson et al., 2017), a result that is more consistent with a global effect on receptor conformation, but future PAMs could conceivably achieve pathway-selective signaling by this mechanism.
An important question relates to the role of membrane lipids in the IC2-region binding site. Roughly half of the site is composed of the cytoplasmic ends of transmembrane helices TM3 and TM4. Virtually all of the amino acids in this area are hydrophobic, and this portion of the binding site is probably covered with lipid tail-groups when not occupied by a D1 PAM. Is there an energetic penalty for displacement of lipids by DETQ, or conversely, is there an energetic advantage for DETQ to insert itself under a blanket of lipids? Does any particular endogenous lipid (e.g., cholesterol) preferentially bind to this region? A full model of the binding site will need to include the influence of the membrane component.
The ability of the D1 IC2 region to retain responsiveness to DETQ when inserted into a b2/D1 chimera suggests several opportunities. Could such a construct be used to obtain a threedimensional structure of the DETQ binding site? Several useful stratagems for stabilizing the agonist state of the b2 receptor are known (Rasmussen et al., 2011a,b). Would the b2/D1 chimera retain responsiveness to DETQ when expressed in a transgenic animal? For instance, such an approach could be used to validate a b2 PAM strategy for treatment of asthma (Ahn et al., 2018) even though no b2 PAMs with in vivo activity have yet been reported. Could the chimera approach be extended to more distant receptors such as adenosine A2a or melanocortin MC4R? The b2/D1 chimera showed a 5-fold loss of potency for DETQ compared with wild-type D1, suggesting at least a mild degree of distortion or strain of the DETQ binding site in the chimera. Would a greater sequence mismatch result in more profound loss of activity, and if so, could stabilizing mutations elsewhere in the receptor be used to restore activity? Chimeric receptors can be a powerful tool (Gearing et al., 2003), but the range of applicability of this approach needs to be explored in more depth.
The D1 PAM CID 2886111 was not affected by any of the IC2 mutations, indicating that it binds to a site that is distinct from the DETQ binding site and that non-local influences such as changes in coupling or receptor expression do not account for the effects of the mutants on the potency of DETQ. In addition, the synergy between CID 2886111 and DETQ confirms that both stabilize the same activated receptor conformation. The overall allosteric boost (a • b 5 g) of DETQ toward CID 2886111 (1.9 • 9.0 5 17-fold) was similar in magnitude to its boost toward dopamine (21 • 1.22 5 25-fold) (Svensson et al., 2017), indicating that DETQ synergizes to a similar degree with both ligands. This observation confirms the prediction from receptor theory (Ehlert, 2005;Kenakin, 2017) that the overall allosteric boost will be the same regardless of whether the compound being potentiated is a partial agonist such as CID 2886111 (in which case the main effect is on efficacy) or a full agonist such as dopamine (in which case the main effect is on affinity).
The general location of the binding site for CID 2886111 could not be definitively established with the D1/D5 chimeras, since its D1 and D5 potency differed by only a factor of three. However, the relative potencies of CID 2886111 for the different chimeras (Fig. 2) tend to favor a site in the C-terminal half of the receptor. One candidate would be the glucagon-like peptide 1 receptor PAM site on the outward-facing parts of the cytoplasmic ends of TMs 5, 6, and 7 (Nolte et al., 2014;Bueno et al., 2016). Additional experiments to identify and characterize the binding site for CID 2886111 are warranted.
How can structural knowledge of allosteric modulator binding sites be used to accelerate drug discovery? In the case of the acyl-tetrahydroisoquinoline series, our original rat/human chimera studies led to the creation of a human D1 receptor knock-in mouse (Svensson et al., 2017), which in turn enabled the animal studies required for advancement of a D1 PAM into clinical trials (http://www.lilly.com/pipeline/). In addition, the D1/D5 chimeras should be useful for locating the binding sites for new D1 PAMs discovered from screening, and single mutants such as W123A and A139M can identify screening hits that are structurally unrelated to the acyl-tetrahydroisoquinolines but nevertheless bind to the same IC2 site. Finally, knowledge of the residues that are important for potency could potentially be used for target hopping to discover allosteric modulators of related receptors. For example, through understanding the residues of the b2 receptor that account for the loss of affinity of DETQ (Fig. 9), it might be possible to remodel the acyltetrahydroisoquinolines into b2 PAMs by introducing compensatory changes into the ligand structure.
In conclusion, this study identifies the IC2 region as the binding site for the D1 PAM DETQ and characterizes the residues that are important for affinity and efficacy. In addition, the results show that the D1 PAM CID 2886111 binds to a different site from DETQ, and that DETQ and CID 2886111 synergize in their effects on cAMP accumulation. These results should aid in the design of novel allosteric modulators of GPCRs.